Research Report Insight Extraction

Problem

When research workflows need answers such as the latest equity research views, sector-level institutional views, ratings, or target prices, the difficult part is turning scattered report fields into a verifiable summary without subjective recommendations. Research Report Insight Extraction constrains these queries to research reports, broker views, institutional ratings, target prices, and analyst views, producing a Markdown result with explicit sources and dates.

How It Works

The skill supports single stock, sector/industry, and multi-stock query modes, using a default one-week window and honoring user-specified ranges. Key steps include:

• Precondition checks: verify IWENCAI_API_KEY and dependent skills such as report-search, hithink-insresearch-query, news-search, and announcement-search.
• Intent parsing: distinguish single stock, sector/industry, and multi-stock requests; pure market or financial queries are out of scope.
• Data retrieval: call report search and institutional research interfaces, preserving institution name, report title, publication date, rating, target price, and core viewpoints.
• Objective synthesis: filter by date, remove duplicates, flag small samples, and output four sections: core views, ratings and target prices, key disagreements, and risk notes, with data source and statistics in the footer.

Boundaries And Caveats

The skill prioritizes no subjective judgment and no fabricated data. Outputs avoid phrasing like "buy", "watchlist", or promotional language; broker names, titles, dates, ratings, and target prices must come from actual returned fields. When institutions disagree, it lists each side rather than reconciling them. If few reports exist, it notes limited sample size; if partial multi-stock queries fail, it delivers successful parts and marks missing results.